More than three billion people now use generative AI tools each month, yet that staggering adoption has not translated into measurable gains in aggregate productivity or employment, according to data gathered by the World Bank. The lag echoes the forty years it took for electricity to appear in productivity statistics, a delay caused not by the technology itself but by the need to reorganise factories and workflows around it. Viewed from Washington, the Federal Reserve confronts a parallel puzzle: even as AI-driven abundance pushes some prices down, the resulting wealth creation will simultaneously drive up the cost of personal services and scarce goods, a dynamic that monetary policy cannot steer.

That reorganisation is proving uneven. A Colombian bank cited by the World Bank found that its most experienced employees resisted AI tools, while Procter & Gamble assembled AI-augmented teams that matched human performance in far less time. Harvard Business School’s Raffaella Sadun describes a J-curve effect: productivity often dips before it rises, as firms absorb the costs of redesigning workflows and reassigning tasks. In the United States, customer-service workers aged 22 to 25 have seen employment fall by 10 per cent since ChatGPT’s 2022 launch, threatening the entry-level roles that build a skilled workforce. Meanwhile, Hong Kong’s education secretary, Christine Choi, has rejected device bans, arguing that students must learn to navigate a digital future rather than be disconnected from it.

Beneath the deployment friction lies an unresolved legal debt. European regulators, under the 2019 Copyright in the Digital Single Market directive and the EU AI Act, now require providers of general-purpose AI models to respect opt-out rights and publish detailed summaries of training data. In the US, a series of lawsuits—including one by authors Andrea Bartz, Charles Graeber and Kirk Wallace Johnson against Anthropic—are testing whether companies must pay for the copyrighted works on which their models were trained. The core tension, as legal observers note, is that the systems now marketed as replacing human labour were built on the creative and scientific output of those same humans.

From Beijing to Silicon Valley, the pace of execution diverges sharply. Chinese firms, visitors report, have already rendered AI invisible by integrating it into everyday payments, transport and communication, treating it as infrastructure rather than innovation. That speed contrasts with a broader philosophical question raised by analysts in Africa and Latin America: whether the relentless pursuit of efficiency risks eroding critical thinking, spiritual life and self-worth, as Pope Leo XIV warned. The next factual milestone to watch is the enforcement of the EU AI Act’s transparency obligations, which will compel model providers to disclose their training sources in sufficient detail—a step that could reshape the economics of the entire sector.